Gait Pattern Generation and Analysis of a Hexapod Ant Robot Using Reinforcement Learning

Abstract

The development of legged robots marks a significant milestone in robotics, offering enhanced capabilities compared to wheeled robots. These robots are engineered to navigate challenging terrains and reach locations inaccessible to humans. Since their inception, legged robots have attracted considerable research and development efforts, leading to the creation of monopods, biped, quadruped, hexapod, and even octopod robots. Among these, hexapod robots have become a preferred choice for researchers due to their superior agility and stability compared to robots with fewer legs, and their lower energy consumption compared to robots with more legs. Despite the advancements in hexapod robots, there remain opportunities for improvement, such as increasing the degrees of freedom to enhance agility and incorporating biomimetic concepts to replicate the naturally optimized locomotion of biological hexapods. This paper presents the development of an ant-inspired kinematic model for a hexapod robot, addressing these improvements. The model is trained using reinforcement learning to achieve an optimal gait pattern, focusing on stability, energy efficiency, and vertical and lateral movements. Additionally, the proposed kinematic model can be further refined through biomimicry, informed by in-depth research on the exoskeleton of ants.